On Sunday 25 January I could see that Clawdbot was getting a lot of social media hype. As people got back to work and tried it out, it completely took over my LinkedIn feed. The project, which started as a tool Austrian developer Peter Steinberger built for his own use, promises to be the "AI that actually does things," from managing calendars to sending messages through your favourite apps.
It’s an open source AI assistant/agent that you can chat to via Whatsapp or Telegram, and carries out tasks autonomously. It’s difficult to recall any one release like this since DeepSeek created shockwaves almost exactly a year ago. Since its release, due to a legal challenge from Anthropic, it’s changed its name to Moltbook and now it’s called OpenClaw.
What can it do?
Manages calendars, sends messages and checks you in for flights autonomously.
Runs locally on your computer rather than in the cloud for privacy.
Supports multiple AI models so users can choose based on security preferences.
Uses downloadable "skill" files to extend functionality and integrations.
Works through chat apps you already use like Slack and WhatsApp.
The good, the bad and the weird
OpenClaw demonstrates what personal AI agents might become: assistants that take action rather than just answer questions.
The project's security limitations show it remains a tool for technical early adopters. Running OpenClaw safely means using a separate server and throwaway accounts, keeping it out of reach for mainstream users attracted by the "AI that does things" promise.
The community has already spawned Moltbook, a social network where AI assistants interact with each other. Andrej Karpathy called this phenomenon "the most incredible sci-fi takeoff-adjacent thing I have seen recently."
Our take
OpenClaw is definitely getting a lot of hype but I until I get time to fully install it and get it running, it’s difficult to know how useful it could be for me. I did start installing it last Monday it’s not that easy. You need to install on terminal and configure API keys and integrations. I’d need to put aside a good hour or two to work through this.
From the many LinkedIn posts I’ve read, I can see its utility, as it can act as an assistant in all sorts of ways - not just doing the tedious admin side of business like clearing inboxes and calendars.
The biggest counter are claims of it being a security nightmare - but you’d expect security professionals to say so. If you hand over your calendars and API keys, isn’t that asking for trouble? While I think considered personal usage is fine, what this release signals to me is a division between corporate control (big companies will absolutely not allow its installation) and entrepreneurial ability to deploy AI agents quickly.
I expect this to be one of the major trends over the next few years - velocity and a blurring of the lines will erode big company positions vs AI first startups.
Another big thing… Anthropic gets sued again
It’s kind of funny that Anthropic position themselves as the good guys when they are getting whacked with another huge lawsuit. Concord Music Group and Universal Music Group are suing the Claude creator, saying it illegally downloaded more than 20,000 copyrighted songs.
This follows the Bartz v. Anthropic case, where a judge ruled that training on copyrighted content is legal but acquiring it through piracy is not. That case cost Anthropic $1.5 billion. Training on pirated material is not okay - it’s bizarre to us that AI companies went down this route, especially Anthropic.
